KVM: x86: Protect the unused bits in MSR exiting flags
authorAaron Lewis <aaronlewis@google.com>
Thu, 14 Jul 2022 16:13:15 +0000 (16:13 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 10 Nov 2022 17:15:25 +0000 (18:15 +0100)
[ Upstream commit cf5029d5dd7cb0aaa53250fa9e389abd231606b3 ]

The flags for KVM_CAP_X86_USER_SPACE_MSR and KVM_X86_SET_MSR_FILTER
have no protection for their unused bits.  Without protection, future
development for these features will be difficult.  Add the protection
needed to make it possible to extend these features in the future.

Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Message-Id: <20220714161314.1715227-1-aaronlewis@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Stable-dep-of: 2e3272bc1790 ("KVM: x86: Copy filter arg outside kvm_vm_ioctl_set_msr_filter()")
Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/x86/kvm/x86.c

index cd22557e2645022644bd23da71ca1a7aab1d8ea2..59c9eb55e6d12e5ceea1c1029367119577ef56ff 100644 (file)
@@ -5770,6 +5770,11 @@ split_irqchip_unlock:
                r = 0;
                break;
        case KVM_CAP_X86_USER_SPACE_MSR:
+               r = -EINVAL;
+               if (cap->args[0] & ~(KVM_MSR_EXIT_REASON_INVAL |
+                                    KVM_MSR_EXIT_REASON_UNKNOWN |
+                                    KVM_MSR_EXIT_REASON_FILTER))
+                       break;
                kvm->arch.user_space_msr_mask = cap->args[0];
                r = 0;
                break;
@@ -5903,6 +5908,9 @@ static int kvm_vm_ioctl_set_msr_filter(struct kvm *kvm, void __user *argp)
        if (copy_from_user(&filter, user_msr_filter, sizeof(filter)))
                return -EFAULT;
 
+       if (filter.flags & ~KVM_MSR_FILTER_DEFAULT_DENY)
+               return -EINVAL;
+
        for (i = 0; i < ARRAY_SIZE(filter.ranges); i++)
                empty &= !filter.ranges[i].nmsrs;